The demographic number is the slowest-moving variable in the file. Getting from 3.9 million graduates to 3.4 million takes sixteen years, which compounds at roughly 0.85% a year [1], and it is barely a forecast at all, since anyone collecting a diploma in 2041 was born around 2023 [9]. That is a planning input. What moves inside a single budget cycle is the price, and at the 57.1% discount rate the National Association of College and University Business Officers found for first-time undergraduates [7], a dollar of sticker tuition lands as 42.9 cents [3], so a school needs about 2.33 enrolled students to replace the revenue one full-payer used to carry [4]. Robert Kelchen's observation that abandoning discounting is not feasible because students will simply choose another college [9] is the mechanism that turns a demand shortfall into a price war rather than a visibly empty campus.
The applications data is where the credit story hides. Common App users applied to 6.56 colleges this cycle against 6.37 last year, a 3.0% increase [6], and since each of those students still enrolls exactly once, the average odds that any given application converts fall from about 15.7% to 15.2% [5]. Crude, or rather the more interesting version of crude, because applications are not admits and the mix differs by school. But it runs the same direction as what the source reports, which is that colleges are having more trouble predicting which acceptees actually show up [4]. For an institution living off net tuition revenue, a forecast error on yield is the default risk.
Meanwhile the fast-moving line item has nothing to do with birth cohorts: 36% fewer F-1 visas were issued in the May-to-August processing window than a year earlier, per State Department data analysed by The Chronicle of Higher Education [14], and international students matter disproportionately because they pay full tuition [16]. Hampshire College, in Amherst, is set to end academic operations after the fall semester [15]. That is not a 2041 event.
Two ways this reads differently. The Philadelphia Fed's estimate of 80 closures between 2025 and 2029 works out to 16 a year, against a pace of at least 17.6 a year implied by more than 300 shutdowns since 2008 [8], so the headline forecast is arguably a mild deceleration rather than a cliff; and the Education Next scenario attributes only 23 additional closures to a 15% five-year enrolment decline [13], which is a modest marginal effect for a large shock. The other read is substitution, not shrinkage: the immediate-enrolment share has fallen from 70% to 62%, an 11.4% relative decline in a decade [10], while public two-year vocational enrolment grew nearly 20% from spring 2020 to spring 2025 [10] against four-year sticker prices approaching or passing $100,000 at dozens of institutions [11].
This is probably wrong, but I would underwrite the price line before the population line. What would falsify it: a discount rate that plateaus and F-1 issuance that recovers while closures cluster anyway. Then the cohort is doing the work, and 500,000 fewer graduates a year [2] is the whole story after all.
